This paper explores finetuning Whisper for transcribing audio from linguistic elicitation of Tira, a Heiban language of Sudan. Audio originates from linguistic fieldwork and is bilingual in English and Tira. We finetune Whisper large-v3 using hand-labeled Tira audio and evaluate the resulting model on bilingual audio. We show that Whisper exhibits catastrophic forgetting of English after only a small amount of training, but that including automatically annotated English spans of audio in the training data dramatically reduces catastrophic forgetting of English while largely preserving ASR performance on monolingual Tira audio. This work is relevant to the study of automatic speech recognition for under-resourced languages and for contexts of bilingualism in a high and low-resourced language.

While there has been substantial progress in text comprehension through simple factoid question answering, more holistic comprehension of a discourse still presents a major challenge (Dunietz et al., 2020). Someone critically reflecting on a text as they read it will pose curiosity-driven, often open-ended questions, which reflect deep understanding of the content and require complex reasoning to answer (Ko et al., 2020; Westera et al., 2020). A key challenge in building and evaluating models for this type of discourse comprehension is the lack of annotated data, especially since collecting answers to such questions requires high cognitive load for annotators.This paper presents a novel paradigm that enables scalable data collection targeting the comprehension of news documents, viewing these questions through the lens of discourse. The resulting corpus, DCQA (Discourse Comprehension by Question Answering), captures both discourse and semantic links between sentences in the form of free-form, open-ended questions. On an evaluation set that we annotated on questions from Ko et al. (2020), we show that DCQA provides valuable supervision for answering open-ended questions. We additionally design pre-training methods utilizing existing question-answering resources, and use synthetic data to accommodate unanswerable questions.